semiconductor laser diodes are the most efficient and compact sources of coherent radiation in near infrared spectral range. As such, they are an obvious choice for use in optical communication systems based on silica fibers which have windows of transparency falling into the emission range of these III-V semiconductor laser diodes. Whereas the basic principle of achieving an optical gain is the same for all laser diodes and relates to an inversion of free carrier population by electrical injection into a forward biased PIN junction, it is the optical cavity solution that mainly differentiates the semiconductor laser designs. One of the most common, and often most preferable, solutions is a distributed feedback (DFB) cavity. It offers a number of advantages to the laser designer, among which include the possibility of providing dynamically single-frequency generation, an ability to control the lasing frequency in the process of fabrication (by means of lithography) and suitability for monolithic integration with optical waveguides, modulators and other waveguide photonic devices. Therefore, semiconductor laser diodes featuring DFB cavities, hereafter referred to simply as DFB lasers, have become the laser sources of choice in advanced optical communication systems. Also, the DFB laser is an on-chip laser source solution for photonic integrated circuits (PICS), which combine different active and passive waveguide-based functional elements of the optical circuit by monolithically integrating them onto one semiconductor substrate.
In the context of the advanced optical networks, these dynamically single-frequency DFB lasers are routinely used all across the network, from local access to long-haul applications, where the system requires low frequency chirp (e.g. high-speed, long-haul propagation over dispersive fiber) or fixed carrier frequency (e.g. dense wavelength division multiplexing) or both. Therefore, DFB lasers are amongst the most demanded and massively produced optical components for telecommunications markets. The great majority of the DFB lasers addressing these markets are designed for operation in either the 1550 nm or 1310 nm wavelength windows, corresponding to the spectral ranges of minimal loss and minimal dispersion of silica fibers, respectively. The material system of choice, which naturally covers these wavelength windows, being indium phosphide (MP) based compound semiconductors from groups III and V, hereafter referred to as III-V semiconductors.
From the PIC prospective, DFB lasers are an attractive design option that not only has the advantage of dynamically single-frequency operation, but also is suitable for a monolithic integration with other waveguide components, both active (e.g. electro-absorption modulator) and passive (e.g. mode converter), onto the same semiconductor substrate, which a laser with a cavity defined by cleaved-facets such as Fabry-Perot (FP) lasers are not. For the majority of applications in today's optical communication systems, PICs featuring laser sources, similar to stand-alone lasers, need to operate in the 1550 nm or 11310 nm wavelength windows, making MP and related semiconductors the natural choices for the substrate and material system of the PIC, respectively.
Despite of a wide-spread manufacturing of the stand-alone InP-based DFB lasers covering the important communication spectral windows above, commercial production of the InP-based PICs having these lasers for the on-chip light sources are still in their infancy and, as of today, practically limited to a few companies. The reason, partially, is in the difficulty of photonic integration itself, but, mainly, an inability to do it in a cost-efficient way, which not only makes it difficult for PICs to compete with micro-optically assembled or hybrid integrated counterparts, but also dramatically narrows market opportunities for those PICs. Therefore, design and fabrication solutions for DFB lasers that would be compatible with a cost-efficient monolithic photonic integration in MP-based material system remain an important problem to solve on the way for advancement of PIC technologies to high-scale commercialization and deployment.
The operating principle of the DFB laser cavity is based on multi-beam interference in an optical waveguide having periodic modulation of the effective index as seen by the guided waves, hereafter referred to as a waveguide Bragg grating. More specifically, a DFB cavity effect occurs when co-directional guided waves, which have experienced multiple reflections from the Bragg grating, are in-phase with each other and add through constructive interference. For any given grating order this happens under conditions of a direct resonant coupling between the contra-directional guided waves occurring near the Bragg wavelength given by Equation (1) below:
                                          λ            B                    =                                    2              ⁢              Λ              ⁢                                                          ⁢                              n                eff                                      m                          ,                            (        1        )            where Λ is the period of the grating and neff is the wavelength dependent effective refractive index of the guided mode. As a result, the DEB cavity is a wavelength selective cavity, with a selection defined by the grating pitch, an advantage that makes DFB lasers attractive for applications requiring emission wavelength control.
Whereas the basic operating principle of the DFB cavity is the same for any form and shape of the Bragg grating, its actual performance depends on many factors including the grating order, type of a spatial modulation of the effective index seen by the guided modes in the optical waveguide bearing the grating, e.g. whether this is a modulation of a real or an imaginary or both the real and imaginary parts of the effective index, and the coupling efficiency of the grating. From this prospective, the choices available to DFB laser designer are many, but in reality the most common solutions are compromises between performance and manufacturability.
Performance-wise, it is always advantageous to have the first order grating, i.e. m=1, and thereby avoiding coupling of the guided waves to, and radiation loss, through non-guided waves, hereafter also referred to as radiative waves. However, for any given emission wavelength λ, the corresponding grating pitch Λ is given by Equation (2) below:
                    Λ        =                              m            ⁢                                                  ⁢            λ                                2            ⁢                          n              eff                                                          (        2        )            and of all the grating orders the shortest grating pitch, and most challenging from the grating patterning point of view, being for m=1. For example, in the wavelength window at 1310 nm, the first order grating pitch should be only ˜200 nm so that for a duty cycle of 50% it requires ˜100 nm lithography line resolution. This is not possible to achieve with a conventional (contact) optical lithography or even a stepper optical lithography usually available to laser fabrication and requires different and more expensive techniques to be used, such as direct electron beam writing, focused ion beam writing, or most commonly optical beam interference.
In the last case, the waveguide Bragg grating is usually formed by fabricating an etched corrugation in a semiconductor layer that is close to the waveguide core as this leads to a strong coupling between the contra-directional guided waves in a spectral range near Bragg resonance and therein leads to favorable DFB cavity characteristics such as narrow resonance line and high rejection of wavelengths outside the resonance range. This technique is well documented in the prior art, see for example S. Akiba et al in U.S. Pat. No. 4,506,367 entitled “Distributed Feedback Semiconductor Laser”, Z-L Liau et al in U.S. Pat. No. 4,722,092 entitled “GaInAsP/InP Distributed Feedback Laser” and G. L. Dolan et al in U.S. Pat. No. 4,701,995 entitled “Method of Making a Non-Planar Buried-Heterostructure Distributed-Feedback Laser” and is nearly universally used in today's commercial fabrication of the DFB lasers operating in the important wavelength ranges, including 1310 nm and 1550 nm communication windows. More recent attempts to improve and/or simplify the process include those reported by T. Takiguchi in U.S. Pat. No. 6,741,630 entitled “Ridge Waveguide Distributed Feedback Laser”, Y. Hisa in U.S. Pat. No. 5,659,562 entitled “Semiconductor Laser including Embedded Diffraction Grating”, and M. Carras et al in U.S. Pat. No. 7,567,606 entitled “Strong Distributed Feedback Semiconductor Laser.”
The fundamental feature of such techniques is that they all require an additional epitaxial growth, performed after the corrugated grating etch, to complete the laser waveguide structure and provide the semiconductor material suitable for forming the electric contact at the top of the laser structure. An overgrowth is a process that not only complicates fabrication and reduces yield by generating defects at the interface of the etched corrugated grating in the proximity to (in a case of the index-coupled grating) or within (in a case of the gain-coupling grating) the laser gain region, but also restricts compatibility of so-processed DFB lasers with other functional elements and thereby limits their usability for monolithic PICs. To overcome this limitation, usually more (selective area) growth steps are required, which follow the (selective area) etching that eliminates unneeded additionally grown semiconductor material in the areas designated to other functional elements of the PIC, which makes the PIC complicated to fabricate while fabrication suffers from yield loss. As a result, deployments of PiCs fabricated by using such multiple growth step techniques have been limited to those where high costs and low yields can be absorbed and volumes relatively low, e.g. large-scale integration PICs for long-haul WDM optical communication systems reported including D. Welch et al in U.S. Pat. No. 7,283,694 entitled “Transmitter Photonic Integrated Circuits (TxPIC) and Optical Transport Networks Employing TxPICs”, C. Joyner in U.S. Pat. No. 7,457,496 entitled “Receiver Photonic Integrated Circuit (RxPIC) Chip Utilizing Compact Wavelength Selective Decombiners”, F. Kish et al in U.S. Pat. No. 7,466,882 entitled “Monolithic Transmitter/Receiver Photonic Integrated Circuit (Tx/Rx PIC)” and R. L. Nagarajan et al in U.S. Pat. No. 7,636,522 entitled “Coolerless Photonic Integrated Circuits (PICs) for WDM Transmission Networks and PICs Operable with a Floating Signal Channel. Grid Changing with Temperature but Fixed Channel Spacing in the Floating Grid.”
For a monolithic PIC technology to win over the multiple hybrid integration technologies commercially deployed in volume, at least in the areas where both can deliver the required functionality and performance even though in different ways, it has to be more cost effective. Therefore, there is a need in the art of photonic integration for developing integration techniques and platforms, which would reduce high fabrication complexity and low-yield processes such as multiple step epitaxial growth, while preserving the required range of the PIC functionality. DEB laser design and fabrication based on defining and etching the corrugated grating in the proximity to or within the laser gain region, with the following up overgrowth, do not seem to meet these requirements and hence other design and fabrication solutions need to be sought in order to make the DFB laser a compatible building block to cost-efficient PICs.
One particular approach to cost-efficient PICs is based on a multi-guide vertical integration (MGVI) platform. This is a generic and versatile technology, which, unlike most of its counterparts, is implementable in one-step epitaxial growth process, e.g. on an lnP substrate, see V. Tolstikhin et al in U.S. Pat. No. 7,444,055 entitled “Integrated Optics Arrangement for Wavelength (De)multiplexing in a Multi-grade Vertical Stack”. In essence, MGVI is a generalization of a twin-guide vertical integration technique, see F. Xia et al in “Photonic Integration using Asymmetric Twin-Waveguide (ATG) Technology (Part 1)—Concepts and Theory” (IEEE J. Sel. Topics in Quant. Electron., 11, 17, 2005) and V. M. Tenon et al in “Photonic Integration using Asymmetric Twin-Waveguide (ATG) Technology (Part 2)—Devices” (IEEE J. Sel. Topics in Quant. Electron., 11, 30, 2005) and the references contained therein, towards multi-functional PICs in which optical waveguides with different functions are vertically stacked in order of ascending guiding layer bandgap wavelength, and adiabatic transitions between them are affected by lateral tapers defined at each guiding level. Functional elements at different vertical levels are optimized independently, while the required PIC functionality is achieved by a proper choice of the guide layers and their relative position in the vertical stack.
Recently, this technique has been successfully demonstrated to be capable of a monolithic integration of all key functions anticipated from PICs for fiber-optics communication systems, including on-chip generation, amplification and detection of light as well as wavelength division multiplexing, mode/spot size converting, beam splitting and re-routing. The design principles, fabrication and characterization of some exemplary MGVI PIC implementation methodologies may be established from references such as V. Tolstikhin et al in U.S. Pat. No. 7,532,784 entitled “Integrated Vertical Wavelength (De)Multiplexer” and Y. Logvin et al in U.S. Pat. No. 7,539,373 entitled “Integrated Lateral Mode Converter” and for actual PICs themselves from references including V. Tolstikhin et al in “Laterally Coupled DFB Lasers for One-Step Growth Photonic Integrated Circuits in InP” (Phot. Tech. Lett. Vol. 21, No. 10, pp 621-623), C. Watson et al in “Optically Pre-Amplified Photodetectors for Multi-Guide Vertical Integration in InP” Paper TuB1.6, International Conte Indium Phosphide and Related Materials 2009), V. Tolstikhin et al in “One Step Growth Optical Transceiver PIC in InP” (ECOC 2009, Paper 8.6.2), S. B. Kuntze et al in “Transmitter and Receiver Solutions for Regrowth Free Multi-Guide Vertical Integration” (Integrated Photonics Research 2010, Paper ITuC5), and K. Pimenov et al in “Analysis of High-Order Surface Etched Gratings for Longitudinal Mode Selection in DFB Lasers” (Numerical Simulation of Optoelectronic Devices 2010, Paper TuC3).
As it concerns an on-chip laser source, the requirement for compatibility with the MGVI platform restricts the design choices by excluding such common laser solutions including cleaved-facet Fabry-Perot cavities (requires physical separation of the laser from the rest of the PIC), distributed reflector/feedback cavities with waveguide Bragg gratings defined at the interface between laser guide layers (these require epitaxial re-growth), butt-coupling to the passive waveguide (these require an additional growth step), and bottom contact at the back surface of the highly conductive substrate (incompatible with monolithic integration). Besides, to preserve the cost efficiency of the MGVI platform, its key differentiating feature in enabling the massive applications in the optical access and interconnect networks as presented by V. Tolstikhin in “Integrated Photonics: Enabling Optical Component Technologies for Next Generation Access Networks” Optical Fiber Communication and Optoelectronics Conference 2007), costly and/or non-volume scalable processes are also undesirable, such as direct-write e-beam lithography.
A solution that addresses all these issues is an effective-ridge waveguide, laterally-coupled grating distributed feedback (LCG-DFB) laser, first reported by Miller et al, as a stand-alone P-contact up device on N+-substrate in GaAs material system in “A Distributed Feedback Ridge Waveguide Quantum Well Heterostructure Laser” (IEEE Phot. Tech. Lett. Vol. 6, No. 9, 1991) as illustrated by first schematic 110 in FIG. 1. This was subsequently re-engineered to be MGVI compatible in an N-contact up device on semi-insulating substrate in InP material system by V. Tolstikhin et al in “Laterally-Coupled DFB Laser for One-Step Growth Photonic Integrated Circuits in InP” (IEEE Phot. Tech. Lett., Vol. 21, pp 621-623, 2009). This design being illustrated in FIG. 2 by second schematic 120 and described below combines the advantages of the DFB laser with its suitability for a regrowth-free monolithic integration with other functional elements, as shown by third schematic 130 in FIG. 1 and described below. The key element of the LCG-DFB laser is the surface etched grating (SEG) formed in the laser's top N-emitter layer by etching two parallel sets of periodic trenches separated by a strip of intact material between them. This SEG provides both lateral confinement, as the average refractive index in each set of periodic trenches is lower than that in the intact material, and the DFB cavity which are processed in one fabrication step. This obviously is an advantage of such a design, which hereafter is referred to as an effective ridge design, the effective ridge being formed by the two parallel sets of trenches and the stripe of intact material between them. MGVI compatible LCG-DFB lasers featuring this design, as a generic building block of multi-functional PIC or a part of a bidirectional transceiver duplex PIC, have been reported by S. B. Kuntze et al in “Transmitter and Receiver Solutions for Regrowth Free Multi-Guide Vertical Integration” (Integrated Photonics Research 2010, Paper ITuC5) and V. Tolstikhin et al in “One Step Growth Optical Transceiver PIC in InP” (ECOC 2009, Paper 8.6.2), respectively. Also, these devices are integral parts of the PIC based transceiver products for access optical networks; see for example OneChip Photonics Inc. (www.onechipphotonics.com/products).
However, the effective ridge DFB cavity has inherent performance limitations associated with the way the SEG is defined and interacts with the guided waves in the laser cavity. More specifically, the coupling efficiency of such a DFB cavity is relatively low, as compared to its counterparts featuring corrugated grating defined within or in a close proximity to the optical waveguide. Since for a given length of the grating, the coupling efficiency defines the laser cavity quality and all performance features associated with it, less efficient coupling requires longer cavities and higher bias currents, which certainly is not a preferable choice in most laser designs and especially those intended for a high-speed direct modulation
A coefficient of direct resonance coupling between contra-propagating guided waves in a waveguide with mth order Bragg grating is given by Equation (3) below:
                              κ          m                =                              (                          π                              λ                ⁢                                                                  ⁢                                  n                  eff                                                      )                    ⁢                      〈                                          φ                *                            |                              A                m                            |              φ                        〉                                              (        3        )            which is defined as the overlap of the normalized guided mode function, φ, with the mth spatial harmonic of the grating Am. In the simple case of a rectangular shaped grating Am is expressed as given below in Equation (4):
                              A          m                =                                            Δ              ⁢                                                          ⁢                              ɛ                ′                                                    m              ⁢                                                          ⁢              π                                ⁢                                          ⁢          sin          ⁢                                          ⁢                      (                          m              ⁢                                                          ⁢              π              ⁢                                                          ⁢              γ                        )                                              (        4        )            where Δε′ is the difference in optical permittivity for a wavelength close to the resonance wavelength λB=2Λneff/m between the etched (grating trench) and unetched (grating tooth) parts of the grating, and γ is the grating duty cycle, defined as a fraction of the grating period occupied by the grating tooth. Accordingly the direct resonance coupling coefficient is reduced to that given below in Equation (5), see G. P. Agrawal and N. K. Dutta, “Long-Wavelength Semiconductor Lasers”, Published by Van Nostrand, 1968:
                              κ          m                =                                            sin              ⁢                                                          ⁢                              (                                  m                  ⁢                                                                          ⁢                  π                  ⁢                                                                          ⁢                  γ                                )                                      m                    ⁢                                    Δ              ⁢                                                          ⁢                              ɛ                ′                                                    λ              ⁢                                                          ⁢                              n                eff                                              ⁢                      〈                                          φ                *                            |                              A                m                            |              φ                        〉                                              (        5        )            
Considering first the term φ*|Am|φ on the right hand side of this equation, which in essence represents the optical mode overlap with the grating; it is relatively small for any grating order since only a tiny fraction of the optical mode, confined vertically in the laser guide underneath the SEG and laterally—by the effective ridge, penetrates the SEG. Such coupling, when wave interaction occurs only due to the evanescent field at the tails of the guided mode, is often referred to as the evanescent field coupling. Second, in a practical implementation, the grating order is limited by the lithography line resolution. So, for a first order grating, which yields the highest coupling coefficient and does not produce radiation losses, in the 1310 nm telecommunications wavelength window, the first order grating requires the line resolution be below 100 nm. Whilst, in principle, this resolution is achievable, e.g. by direct-write e-beam or X-line stepper lithography, practically it is out of reach in most cases. For the examples above, the former is too slow a process for commercial production while the latter is a very expensive tool, usually not found in III-V optoelectronic fabrication facilities. More typical for such facilities, which, opposite to their silicon counterparts do not enjoy such high volumes that would justify expensive deep ultra-violet stepper lithography, yet not quite a commodity, is a I-line stepper with an emission wavelength of 362 nm. Accordingly, for the example of the 1310 nm wavelength range the first order grating is beyond diffraction limit of such a stepper, and the only available solutions for the stepper lithography are higher order gratings, presumably m≥3. But increasing the grating order further reduces the resonance coupling coefficient, by a factor of 1/m, thereby making an evanescent-field coupled LCG-DFB even a more challenging task.
Therefore, there is an inherent problem in achieving high coupling efficiency in LCG-DFB lasers, which, otherwise, are very attractive from their design and fabrication points of view and also well suited for a regrowth-free monolithic integration with other functional elements into MGVI based PICs for low-cost transceiver applications, such as optical access networks and interconnects. Within the prior art, the research has sought a solution to enhancement of the coupling efficiency through various advanced—and more complicated than a straightforward surface grating definition and etch—techniques. For example such techniques have included forming the grating, either by optical stepper lithography (see for example Reid et al in “Narrow Linewidth and. High Power Distributed Feedback Lasers Fabricated without a Regrowth Step”, Proc. ECOC 2003 Rimini) or electron beam lithography (e.g. D. K. Oh et al in US Patent Application 2006/0120428 entitled “Distributed Feedback (DFB) Semiconductor Laser and Fabrication Method Thereof”) in the entire outer profile of the semiconductor waveguide structure;) or forming grating elements in sloped lateral sidewalk of the waveguide structure (see for example Y. T. Lee et al in US Patent Application 2008/0261157 entitled “Semiconductor Laser Device and Method of Manufacturing the Same”). Also, approaches that include forming buried lateral grating elements, and hence require a re-growth step, have been proposed (see for example Y. Watanabe et al in US Patent Application 2001/0046246 entitled “Ridge Type Semiconductor Laser of Laterally Coupled Distributed Feedback and Method Manufacturing the Same” and J. R. Reithmaier et al in US Patent Applications 2006/0172446 and 2009/0117678 each entitled “Semiconductor Laser with a Weakly Coupled Grating.”). All these or similar techniques, while enhancing the coupling efficiency, also complicate fabrication or/and require additional growth step(s). However, it would be highly advantageous to find a solution to providing DFBs within PICS that would enhance the coupling efficiency while preserving a simple and straightforward SEG fabrication process, compatible with the MGVI platform. Since the fabrication process change is not an option, such a solution should be sought in a design change, obviously.
An approach that enhances the coupling efficiency of the laterally coupled SEG (LC-SEG) within the context of a MGVI manufacturing methodology has been proposed by V. Tolstikhin et al in U.S. Pat. No. 7,609,919 entitled “Coupling-Enhanced Surface Etched Gratings” and employs additional effective ridge waveguide elements on either side of the active effective ridge waveguide, such that the side effective ridges provide an intended dilution of the lateral confinement of the optical mode resulting in an increased overlap of the this mode with the SEG regions forming the active and side effective ridges. In effect, the coupling efficiency of the grating can be increased, but only if the effective ridges are narrow enough, as well as the LC-SEG areas that separate them—narrower than the wavelength in the semiconductor material, which is difficult to implement because of the lithography resolution limitations. Another approach to enhancing the coupling efficiency of the LC-SEG, which also exploits the idea of the lateral mode dilution, even though in a different way, has been described by C. Watson et al in U.S. Pat. No. 7,796,656 entitled “Enhanced Efficiency Laterally-Coupled Distributed Feedback Laser.” Within this prior art, the LC-SEG DEB laser operates in the first order lateral mode, as opposed to the conventional fundamental zero order mode, thereby having the optical mode field mostly aside from the active effective ridge, where it is better overlapped with the LC-SEG and hence provides higher coupling with the LC-SEG. This design has yet another advantage, in that it reduces the mode overlap with the metal contact atop the active effective ridge and impact of the SEG etch rounding/irregularities at the trench edges next to the active effective ridge, thereby reducing intracavity loss and further improving the coupling efficiency. On a flip side, this laser needs to be complemented with a certain waveguide arrangement that transfers the first order lateral mode, in which the laser operates, into the zero order optical mode, in which optical signals can be coupled into a fiber or other parts of the PIC. Whereas such an arrangement is feasible within the MGVI platform, e.g. in a form of the lateral mode converter described by Y. Logvin et al in U.S. Pat. No. 7,539,373 B1 entitled “Integrated Lateral Mode Converter”, still it complicates fabrication and also introduces additional insertion loss.
Therefore, the problem of designing a SEG-DFB laser that is not only compatible with MGVI platform, but also is economic in terms of its fabrication within this platform, while providing high coupling efficiency with the SEG, remains largely unsolved in the prior art. It would be beneficial if such a design allowed for the coupling efficiency enhancement without compromising the optical mode confinement within the gain region, something that the prior art teaching does not allow since it taught the optical mode dilution as means for the coupling enhancement with the SEG. High optical mode confinement within the gain region being important for reduction of threshold current for lasing operation and temperature sensitivity of the laser. It would be further beneficial if the SEG-DFB laser design relaxed the fabrication tolerances, e.g. by avoiding the optical mode overlap with the parts of the SEG that are most difficult to control in a process of fabrication, notably the SEG trench edges at the effective ridge side. Corner rounding and irregularities that occur in these trenches just aside from the effective ridge may badly impact the coupling efficiency and scattering loss when overlap with the optical mode, as it is happening in LC-SEG designs taught by the prior art. It is an intention for the invention to overcome the above-mentioned limitations in the prior art.